Batch Processing Support for Model Inference Using Tabular Data

[thuongle2210]

hi @habedi,
I see Infera's roadmap supports:
Batch Processing
Inference on batches for models with dynamic dimensions. => Already supported.
Automatic batch splitting for models with fixed batch size. => Not yet supported.
I see the maintainer currently only supports input as blob. I propose the feature with input like this:

Data preparation:
create or replace table batch_raw_data as select 1 as id, 1.0::float as f1, 2.0::float as f2, 3.0::float as f3 union all select 1 as id, 1.0::float as f1, 2.0::float as f2, 4.0::float as f3;

then

select infera_predict('linear_model', f1, f2, f3) from batch_raw_data;

Expected output is two rows of scores.